1 hack.txt for Citadel/UX
2 (possibly a little out of date)
4 Much of this document is borrowed from the original hack.doc from
5 Citadel-CP/M and Citadel-86, because many of the concepts are the same. Hats
6 off to whoever wrote the original, for a fine document that inspired the
7 implementation of Citadel for Unix.
9 Note that this document is really out of date. It doesn't cover anything
10 about the threaded server architecture or any of the network stuff. What is
11 covered here is the basic architecture of the databases.
13 But enough of the preamble. Here's how Citadel/UX works :)
15 Here are the major databases to be discussed:
17 msgmain The big circular file that contains message text
18 quickroom Contains room info such as room names, stats, etc.
19 fullroom One fullrm file per room: message numbers and pointers.
20 usersupp Contains info for each user on the system.
22 The fundamental structure of the system differs greatly from the way
23 Citadels used to work. Citadel now depends on a record manager or database
24 manager of some sort. Thanks to the API which is in place for connecting to
25 a data store, any record manager may be used as long as it supports the
26 storage and retrieval of large binary objects (blobs) indexed by unique keys.
27 Please see database.c for more information on data store primitives.
29 The message base (MSGMAIN) is a big file of messages indexed by the message
30 number. Messages are numbered consecutively and start with an FF (hex)
31 byte. Except for this FF start-of-message byte, all bytes in the message
32 file have the high bit set to 0. This means that in principle it is
33 trivial to scan through the message file and locate message N if it
34 exists, or return error. (Complexities, as usual, crop up when we
35 try for efficiency...)
37 Each room is basically just a list of message numbers. Each time
38 we enter a new message in a room, its message number is appended to the end
39 of the list. If an old message is to be expired, we must delete it from the
40 message base. Reading a room is just a matter of looking up the messages
41 one by one and sending them to the client for display, printing, or whatever.
43 Implementing the "new message" function is also trivial in principle:
44 we just keep track, for each caller in the userlog, of the highest-numbered
45 message which existed on the *last* call. (Remember, message numbers are
46 simply assigned sequentially each time a message is created. This
47 sequence is global to the entire system, not local within a room.) If
48 we ignore all message-numbers in the room less than this, only new messages
49 will be printed. Voila!
52 Message format on disk (MSGMAIN)
54 As discussed above, each message begins with an FF byte.
56 The next byte denotes whether this is an anonymous message. The codes
57 available are MES_NORMAL, MES_ANON, or MES_AN2 (defined in citadel.h).
59 The third byte is a "message type" code. The following codes are defined:
60 0 - "Traditional" Citadel format. Message is to be displayed "formatted."
61 1 - Plain pre-formatted ASCII text (otherwise known as text/plain)
62 4 - MIME formatted message. The text of the message which follows is
63 expected to begin with a "Content-type:" header.
65 After these three opening bytes, the remainder of
66 the message consists of a sequence of character strings. Each string
67 begins with a type byte indicating the meaning of the string and is
68 ended with a null. All strings are printable ASCII: in particular,
69 all numbers are in ASCII rather than binary. This is for simplicity,
70 both in implementing the system and in implementing other code to
71 work with the system. For instance, a database driven off Citadel archives
72 can do wildcard matching without worrying about unpacking binary data such
73 as message ID's first. To provide later downward compatability
74 all software should be written to IGNORE fields not currently defined.
76 The type bytes currently defined are:
78 BYTE Mnemonic Comments
80 A Author Name of originator of message.
81 B Phone number The dialup number of the system this message
82 originated on. This is optional, and is only
83 defined for helping implement C86Net gateways.
84 D Destination Contains name of the system this message should
85 be sent to, for mail routing (private mail only).
86 E Extended ID A persistent alphanumeric Message ID used for
87 network replication. When a message arrives that
88 contains an Extended ID, any existing messages which
89 contain the same Extended ID and are *older* than this
90 message should be deleted. If there exist any messages
91 with the same Extended ID that are *newer*, then this
92 message should be dropped.
93 G Gateway domain This field is provided solely for the implementation
94 of C86Net gateways, and holds the C86Net domain of
95 the system this message originated on. Unless you're
96 implementing such a gateway, there's no need to even
97 bother with this field.
98 H HumanNodeName Human-readable name of system message originated on.
99 I Original ID A 32-bit integer containing the message ID on the
100 system the message *originated* on.
101 M Message Text Normal ASCII, newlines seperated by CR's or LF's,
102 null terminated as always.
103 N Nodename Contains node name of system message originated on.
104 O Room Room of origin.
105 P Path Complete path of message, as in the UseNet news
106 standard. A user should be able to send Internet mail
107 to this path. (Note that your system name will not be
108 tacked onto this until you're sending the message to
110 R Recipient Only present in Mail messages.
111 S Special field Only meaningful for messages being spooled over a
112 network. Usually means that the message isn't really
113 a message, but rather some other network function:
114 -> "S" followed by "FILE" (followed by a null, of
115 course) means that the message text is actually an
116 IGnet/Open file transfer.
117 T Date/Time A 32-bit integer containing the date and time of
118 the message in standard UNIX format (the number
119 of seconds since January 1, 1970 GMT).
120 U Subject Optional. Developers may choose whether they wish to
121 generate or display subject fields. Citadel/UX does
122 not generate them, but it does print them when found.
126 Let <FF> be a 0xFF byte, and <0> be a null (0x00) byte. Then a message
129 Apr 12, 1988 23:16 From Test User In Network Test> @lifesys (Life BBS)
132 might be stored as...
133 <FF><40><0>I12345<0>Pneighbor!lifesys!test_user<0>T576918988<0> (continued)
134 -----------|Mesg ID#|--Message Path---------------|--Date------
136 AThe Test User<0>ONetwork Test<0>Nlifesys<0>HLife BBS<0>MHave a nice day!<0>
137 |-----Author-----|-Room name-----|-nodename-|Human Name-|--Message text-----
139 Weird things can happen if fields are missing, especially if you use the
140 networker. But basically, the date, author, room, and nodename may be in any
141 order. But the leading fields and the message text must remain in the same
142 place. The H field looks better when it is placed immediately after the N
147 Citadel nodes network by sharing one or more rooms. Any Citadel node
148 can choose to share messages with any other Citadel node, through the sending
149 of spool files. The sending system takes all messages it hasn't sent yet, and
150 spools them to the recieving system, which posts them in the rooms.
152 Complexities arise primarily from the possibility of densely connected
153 networks: one does not wish to accumulate multiple copies of a given
154 message, which can easily happen. Nor does one want to see old messages
155 percolating indefinitely through the system.
157 This problem is handled by keeping track of the path a message has taken over
158 the network, like the UseNet news system does. When a system sends out a
159 message, it adds its own name to the bang-path in the <P> field of the
160 message. If no path field is present, it generates one.
162 With the path present, all the networker has to do to assure that it doesn't
163 send another system a message it's already received is check the <P>ath field
164 for that system's name somewhere in the bang path. If it's present, the system
165 has already seen the message, so we don't send it. (Note that the current
166 implementation does not allow for "loops" in the network -- if you build your
167 net this way you will see lots of duplicate messages.)
169 The above discussion should make the function of the fields reasonably clear:
171 o Travelling messages need to carry original message-id, system of origin,
172 date of origin, author, and path with them, to keep reproduction and
173 cycling under control.
175 (Uncoincidentally) the format used to transmit messages for networking
176 purposes is precisely that used on disk, except that there may be any amount
177 of garbage between the null ending a message and the <FF> starting the next
178 one. This allows greater compatibility if slight problems crop up. The current
179 distribution includes netproc.c, which is basically a database replicator;
180 please see network.txt on its operation and functionality (if any).
184 At this point, all hardware-dependent stuff has been removed from the
185 system. On the server side, most of the OS-dependent stuff has been isolated
186 into the sysdep.c source module. The server should compile on any POSIX
187 compliant system with a full pthreads implementation and TCP/IP support. In
188 the future, we may try to port it to non-POSIX systems as well.
190 On the client side, it's also POSIX compliant. The client even seems to
191 build ok on non-POSIX systems with porting libraries (such as the Cygnus
195 "Room" records (quickroom)
197 The rooms are basically indices into msgmain, the message database.
198 As noted in the overview, each is essentially an array of pointers into
199 the message file. The pointers consist of a 32-bit message ID number
200 (we will wrap around at 32 bits for these purposes).
202 Since messages are numbered sequentially, the
203 set of messages existing in msgmain will always form a continuous
204 sequence at any given time.
206 That should be enough background to tackle a full-scale room. From citadel.h:
209 char QRname[20]; /* Max. len is 19, plus null term */
210 char QRpasswd[10]; /* Only valid if it's a private rm */
211 long QRroomaide; /* User number of room aide */
212 long QRhighest; /* Highest message NUMBER in room */
213 long QRgen; /* Generation number of room */
214 unsigned QRflags; /* See flag values below */
215 char QRdirname[15]; /* Directory name, if applicable */
216 char QRfloor; /* (not yet implemented) */
219 #define QR_BUSY 1 /* Room is being updated, WAIT */
220 #define QR_INUSE 2 /* Set if in use, clear if avail */
221 #define QR_PRIVATE 4 /* Set for any type of private room */
222 #define QR_PASSWORDED 8 /* Set if there's a password too */
223 #define QR_GUESSNAME 16 /* Set if it's a guessname room */
224 #define QR_DIRECTORY 32 /* Directory room */
225 #define QR_UPLOAD 64 /* Allowed to upload */
226 #define QR_DOWNLOAD 128 /* Allowed to download */
227 #define QR_VISDIR 256 /* Visible directory */
228 #define QR_ANONONLY 512 /* Anonymous-Only room */
229 #define QR_ANON2 1024 /* Anonymous-Option room */
230 #define QR_NETWORK 2048 /* Shared network room */
231 #define QR_PREFONLY 4096 /* Preferred users only */
233 [Note that all components start with "QR" for quickroom, to make sure we
234 don't accidentally use an offset in the wrong structure. Be very careful
235 also to get a meaningful sequence of components --
236 some C compilers don't check this sort of stuff either.]
238 QRgen handles the problem of rooms which have died and been reborn
239 under another name. This will be clearer when we get to the userlog.
240 For now, just note that each room has a generation number which is
241 bumped by one each time it is recycled.
243 QRflags is just a bag of bits recording the status of the room. The
246 QR_BUSY This is to insure that two processes don't update the same
247 record at the same time, even though this hasn't been
249 QR_INUSE 1 if the room is valid, 0 if it is free for re-assignment.
250 QR_PRIVATE 1 if the room is not visible by default, 0 for public.
251 QR_PASSWORDED 1 if entry to the room requires a password.
252 QR_GUESSNAME 1 if the room can be reached by guessing the name.
253 QR_DIRECTORY 1 if the room is a window onto some disk/userspace, else 0.
254 QR_UPLOAD 1 if users can upload into this room, else 0.
255 QR_DOWNLOAD 1 if users can download from this room, else 0.
256 QR_VISDIR 1 if users are allowed to read the directory, else 0.
257 QR_ANONONLY 1 if all messages are to recieve the "****" anon header.
258 QR_ANON2 1 if the user will be asked if he/she wants an anon message.
259 QR_NETWORK 1 if this room is shared on a network, else 0.
260 QR_PREFONLY 1 if the room is only accessible to preferred users, else 0.
262 QRname is just an ASCII string (null-terminated, like all strings)
263 giving the name of the room.
265 QRdirname is meaningful only in QR_DIRECTORY rooms, in which case
266 it gives the directory name to window.
268 QRpasswd is the room's password, if it's a QR_PASSWORDED room. Note that
269 if QR_PASSWORDED or QR_GUESSNAME are set, you MUST also set QR_PRIVATE.
270 QR_PRIVATE by itself designates invitation-only. Do not EVER set all three
271 flags at the same time.
273 QRroomaide is the user number of the room's room-aide (or zero if the room
274 doesn't have a room aide). Note that if a user is deleted, his/her user number
275 is never used again, so you don't have to worry about a new user getting the
276 same user number and accidentally becoming a room-aide of one or more rooms.
278 The only field new to us in quickroom is QRhighest, recording the
279 most recent message in the room. When we are searching for rooms with
280 messages a given caller hasn't seen, we can check this number
281 and avoid a whole lot of extra disk accesses.
283 There used to also be a structure called "fullroom" which resided in one
284 file for each room on the system. This has been abandoned in favour of
285 "message lists" which are variable sized and simply contain zero or more
286 message numbers. The message numbers, in turn, point to messages on disk.
288 User records (usersupp)
290 This is the fun one. Get some fresh air and plug in your thinking cap
291 first. (Time, space and complexity are the eternal software rivals.
292 We've got lots of log entries times lots of messages spread over up to nnn
293 rooms to worry about, and with multitasking, disk access time is important...
294 so perforce, we opt for complexity to keep time and space in bounds.)
296 To understand what is happening in the log code takes a little persistence.
297 You also have to disentangle the different activities going on and
298 tackle them one by one.
300 o We want to remember some random things such as terminal screen
301 size, and automatically set them up for each caller at login.
303 o We want to be able to locate all new messages, and only new
304 messages, efficiently. Messages should stay new even if it
305 takes a caller a couple of calls to get around to them.
307 o We want to remember which private rooms a given caller knows
308 about, and treat them as normal rooms. This means mostly
309 automatically seeking out those with new messages. (Obviously,
310 we >don't< want to do this for unknown private rooms!) This
311 has to be secure against the periodic recycling of rooms
314 o We want to support private mail to a caller.
316 o We want to provide some protection of this information (via
317 passwords at login) and some assurance that messages are from
318 who they purport to be from (within the system -- one shouldn't
319 be able to forge messages from established users).
321 Lifting another page from citadel.h gives us:
323 struct usersupp { /* User record */
324 int USuid; /* uid account is logged in under */
325 char password[20]; /* password */
326 long lastseen[MAXROOMS]; /* Last message seen in each room */
327 char generation[MAXROOMS]; /* Generation # (for private rooms) */
328 char forget[MAXROOMS]; /* Forgotten generation number */
329 unsigned flags; /* See US_ flags below */
330 int screenwidth; /* For formatting messages */
331 int timescalled; /* Total number of logins */
332 int posted; /* Number of messages posted (ever) */
333 char fullname[26]; /* Bulletin Board name for messages */
334 char axlevel; /* Access level */
335 long usernum; /* Eternal user number */
336 long lastcall; /* Last time the user called */
339 #define US_PERM 1 /* Permanent user; don't scroll off */
340 #define US_LASTOLD 16 /* Print last old message with new */
341 #define US_EXPERT 32 /* Experienced user */
342 #define US_UNLISTED 64 /* Unlisted userlog entry */
343 #define US_NOPROMPT 128 /* Don't prompt after each message */
344 #define US_PREF 1024 /* Preferred user */
346 Looks simple enough, doesn't it? One topic at a time:
348 Random configuration parameters:
349 -screenwidth is the caller's screen width. We format all messages to this
350 width, as best we can. flags is another bit-bag, recording whether we want
351 prompts, people who want to suppress the little automatic hints all through
354 Attachments, names & numbers:
355 -USuid is the uid the account was established under. For most users it will
356 be the same as BBSUID, but it won't be for users that logged in from the shell.
357 -fullname is the user's full login name.
358 -usernum is the user's ID number. It is unique to the entire system:
359 once someone has a user number, it is never used again after the user is
360 deleted. This allows an easy way to numerically represent people.
361 -password is the user's password.
362 -axlevel is the user's access level, so we know who's an Aide, who's a problem
363 user, etc. These are defined and listed in the system.
366 -timescalled is the number of times the user has called.
367 -posted is the number of messages the user has posted, public or private.
370 -lastcall holds the date and time (standard Unix format) the user called, so
371 we can purge people who haven't called in a given amount of time.
373 Finding new messages:
374 This is the most important. Thus, it winds up being the most
375 elaborate. Conceptually, what we would like to do is mark each
376 message with a bit after our caller has read it, so we can avoid
377 printing it out again next call. Unfortunately, with lots of user
378 entries this would require adding lots of bits to each message... and
379 we'd wind up reading off disk lots of messages which would never
380 get printed. So we resort to approximation and a small table.
382 The approximation comes in doing things at the granularity of
383 rooms rather than messages. Messages in a given room are "new"
384 until we visit it, and "old" after we leave the room... whether
385 we read any of them or not. This can actually be defended: anyone
386 who passes through a room without reading the contents probably just
387 isn't interested in the topic, and would just as soon not be dragged
388 back every visit and forced to read them. Given that messages are
389 numbered sequentially, we can simply record the most recent message ID#
390 of each room as of the last time we visited it. Very simple.
392 Putting it all together, we can now compute whether a given room
393 has new messages for our current caller without going to the message base
394 index (fullroom) at all:
396 > We get the usersupp.lastseen[] for the room in question
397 > We compare this with the room's quickroom.QRhighest, which tells us
398 what the most recent message in the room is currently.
401 REMEMBERING WHICH PRIVATE ROOMS TO VISIT
403 This looks trivial at first glance -- just record one bit per room per
404 caller in the log records. The problem is that rooms get recycled
405 periodically, and we'd rather not run through all the log entries each
406 time we do it. So we adopt a kludge which should work 99% of the time.
408 As previously noted, each room has a generation number, which is bumped
409 by one each time it is recycled. As not noted, this generation number
410 runs from 0 -> 127 (and then wraps around and starts over).
411 When someone visits a room, we set usersupp.generation for the room
412 equal to that of the room. This flags the room as being available.
413 If the room gets recycled, on our next visit the two generation numbers
414 will no longer match, and the room will no longer be available -- just
415 the result we're looking for. (Naturally, if a room is public,
416 all this stuff is irrelevant.)
418 This leaves only the problem of an accidental matchup between the two
419 numbers giving someone access to a Forbidden Room. We can't eliminate
420 this danger completely, but it can be reduced to insignificance for
421 most purposes. (Just don't bet megabucks on the security of this system!)
422 Each time someone logs in, we set all "wrong" generation numbers to -1.
423 So the room must be recycled 127 times before an accidental matchup
424 can be achieved. (We do this for all rooms, INUSE or dead, public
425 or private, since any of them may be reincarnated as a Forbidden Room.)
427 Thus, for someone to accidentally be led to a Forbidden Room, they
428 must establish an account on the system, then not call until some room
429 has been recycled 127 to 128 times, which room must be
430 reincarnated as a Forbidden Room, which someone must now call back
431 (having not scrolled off the userlog in the mean time) and read new
432 messages. The last clause is about the only probable one in the sequence.
433 The danger of this is much less than the danger that someone will
434 simply guess the name of the room outright (if it's a guess-name room)
435 or some other human loophole.
439 This is exactly the opposite of private rooms. When a user chooses to
440 forget a room, we put the room's generation number in usersupp.forget for
441 that room. When doing a <K>nown rooms list or a <G>oto, any matchups cause
442 the room to be skipped. Very simple.
444 SUPPORTING PRIVATE MAIL
446 Can one have an elegant kludge? This must come pretty close.
448 Private mail is sent and recieved in the Mail> room, which otherwise
449 behaves pretty much as any other room. To make this work, we have a
450 separate Mail> room for each user behind the scenes. The actual room name
451 in the database looks like "0000001234.Mail" (where '1234' is the user
452 number) and it's flagged with the QR_MAILBOX flag. The user number is
453 stripped off by the server before the name is presented to the client.
455 This requires a little fiddling to get things just right. For example,
456 make_message() has to be kludged to ask for the name of the recipient
457 of the message whenever a message is entered in Mail>. But basically
458 it works pretty well, keeping the code and user interface simple and
462 PASSWORDS AND NAME VALIDATION
464 This has changed a couple of times over the course of Citadel's history. At
465 this point it's very simple, again due to the fact that record managers are
466 used for everything. The user file (usersupp) is indexed using the user's
467 name, converted to all lower-case. Searching for a user, then, is easy. We
468 just lowercase the name we're looking for and query the database. If no
469 match is found, it is assumed that the user does not exist.
471 This makes it difficult to forge messages from an existing user. (Fine
472 point: nonprinting characters are converted to printing characters, and
473 leading, trailing, and double blanks are deleted.)